Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
This could go in the pets section but it applies to all of us, beast or human.
I just had our kitty River at the vet's where they did the normal labs. And normally the labs are fairly easy to read, often there's a key and a good doc will point out items and explain them.
That didn't happen. The labs look like a kid put them together, there are some bar graphs that do not correspond to tests, they're just out there on the right side of the sheet.
Page three is maybe 10 more bar graphs with nothing to explain them. Not a word, not an abbreviation. Just graphs.
I may go back to this vet but it will only be to try to get an interpretation. A friend's husband has been a lab guy, high up, for decades and she said he'd have a gander and see what he sees.
But I thought, too bad there's not a bot program somewhere. I could scan these in and maybe get an interpretation. She said not to their knowledge.
That would seem like a field ripe for development and I'm surprised there's nothing like that out there.
Waiting on lab doctor's interpretation, meanwhile.
That isn't the way historically that labs are interpreted.
Let's start off with the most obvious gross inaccuracy and that is lab error where samples get switched. One is simply going by the numbers on a form and diagnosing or treating solely based on the numbers. That is just dangerous.
Let's take the next example and that is testing inaccuracies based on technical limitations. There is no absolute demarcation between the normal population and disease populations. Most of the time there is a slow graduated increase or decrease that one can find both normal populations and disease populations having the same numbers. Gross numbers outside the reference range are more strongly associated with disease but they also lack sensitivity in detecting disease early on.
If one increases sensitivity for detecting disease fearing missing any disease state then the threshold is lowered and you will find normal people being called abnormal. That is a false positive. If you increase the threshold number to increase specificity then you end up missing people with early disease being called as negative thus a false negative.
In order to not misinterpret the results all results must be interpreted within a clinical context that involves history, signs and symptoms. That is obviously problematic with animals where they can't talk.
All of those variables and more will cause problems and AI won't solve those problems. A computer program is usually symptoms based involving the differential diagnosis at the top. You can start off with algorithms involving abnormal tests at the top but it is less accurate. The reason for that is as mentioned the same result can have different meanings and requires other input to fully interpret.
After saying that as a professional doing the work I would or should say was forced into making objective calls when doing lab work. I would consider minimal clinical information provided but tried to be objective when making decisions on laboratory findings.
The final decision thus includes clinical information and objective laboratory findings which leads the way towards a diagnosis.
Please register to participate in our discussions with 2 million other members - it's free and quick! Some forums can only be seen by registered members. After you create your account, you'll be able to customize options and access all our 15,000 new posts/day with fewer ads.
